The accuracy of what you're opposing is very very low. Calculate 10 rotations per minute. This would mean at least 1 full minute to get any result. I would suggest, if you can't use J.Guys idea (which I'd say is briliant) then consider this:
Place an optical interruptor, an encarved (or neatly drilled) disk on the driving wheel and you get a much better result. Something like a mouse construction. If the disk is not possible then maybe a bar-code on the wheel and a reflective opto-couppler could do the trick. This way you could have, say 300 pulses per minute.
Have your program scan the time-interval between two pulses and there you are.